Files in this item
|(no description provided)|
|Title:||A Natural Language Interface for Retrieval of Information From a Dairy Database|
|Author(s):||Jones, Lawrence Raymond|
|Doctoral Committee Chair(s):||Spahr, Sidney L.|
|Department / Program:||Animal Science|
|Degree Granting Institution:||University of Illinois at Urbana-Champaign|
|Abstract:||A natural-language interface (NLI) was developed in LISP for use with IBM-AT-compatible microcomputers to provide flexible access to computerized dairy records. NLI was designed to retrieve information from a dBASE III format database using the University of Illinois Dairybase program as the developmental database. NLI could answer domain specific queries concerning information about individual animals or specific subgroups of the herd. It could also summarize statistics about the herd or subgroups within the herd.
Users interacted with NLI using an input module which provided for use of abbreviations and automatic correction of misspelled words. Interpretation of input began by determining the syntactic structure of the request using the augmented transition network (ATN) approach to parsing. The ATN was implemented using LISP functions as continuations to achieve chronological backtracking. These functions were defined using a specially-designed high-level macro language to facilitate modification. Syntactic information of the lexicon was stored as property lists on LISP symbols. NLI proceeded by converting syntactic structures to semantic concepts representing underlying notions of the input using semantic rules, implemented as LISP functions. Resulting semantic concepts were represented using a knowledge-based object-oriented paradigm. The concepts contained details for retrieving and printing information associated with the concept; this new semantic approach for answering questions has been coined "conceptual semantics". Answers were printed in a natural-language format with additional pertinent information needed to make a decision. Most answers could also be presented in a tabular format containing only requested information.
NLI required 623 kilobytes of extended memory to be loaded. Memory needed for operation was determined by that needed to syntactically analyze input. Almost all requests were successfully analyzed when five megabytes of memory was available. Time to respond to queries was dependent on the type of query and database size. Simple queries were generally analyzed in less than four seconds; complex queries generally required less than eight seconds. Retrieval of information from the database and formulation of the answer required from one second to five minutes depending on the number of animals in the request and database size.
Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 1988.
|Date Available in IDEALS:||2014-12-16|